For AI agents: the complete documentation index is available at /llms.txt, the full documentation bundle is available at /llms-full.txt, and this page is available as Markdown at /changelog/2026-07-15.md.
2026-07-15
New
Computer detail shows Runtime availability, capabilities, and how many Agents use each Computer.
Runtimes can be rescanned and diagnosed when the local environment changes.
Task-list and board tabs and filters are kept in the URL, so refresh and shared links restore the view.
Board cards can show live dependency, blocked progress, and completion state.
Related-task search stays inside the current Space.
The new-task entry remembers the last Space you used.
Space members can be assigned Administrator, Member, or Viewer roles. These roles apply only within a Space and complement Organization roles.
Task lists support task-type tabs and filters. In cross-Space filters, identical task-type names are shown as one option.
Improvements
Computer detail, Runtime management, hover info, and action menus are clearer.
Filters, paging, and counts behave the same across the task list, board, and My tasks.
Task-type management links, long status labels, and sidebar active states are easier to follow.
Get started steps are clearer and have a dedicated page.
Connecting a Computer and creating an Agent are smoother, including duplicate-name checks.
Outdated onboarding steps are removed.
Private Jira URLs and HTTP / HTTPS handling are clearer, with better connection-error messages.
Empty import history is hidden when there are no records.
Join-Space dialog copy and interaction are cleaner.
Agent visibility and execution permissions are easier to understand.
Retry-run audit records keep the right actor.
First-time Claude and Codex startup is faster and more reliable for new users.
Queued Agent work is shown consistently in task lists, boards, Inbox, and task details, then updates when execution starts.
Fixes
Board counts, filter results, and summary totals are accurate.
Labels imported from Jira belong to the right Space.
Agents can read attachments, and images in task descriptions preview instead of 404.
Restored filters, paging cache, and multi-filter summaries behave correctly.
Computer Runtime and Agent counts display correctly.